Project Agorá real-value testing and the rulebook agenda

In July 2026, 28 financial institutions and central banks from Asia, Europe and North America settled 30 transactions worth about CHF 800,000 using tokens with real monetary value. The five central banks that took part in the test were the Bank of England, the Bank of France (on behalf of the Eurosystem), the Bank of Japan, the Bank of Korea and the Swiss National Bank. The Bank of Korea, together with NongHyup Bank and Shinhan Bank, tested an interbank won transfer of KRW 20 million using tokenized central bank reserves.

Project Agorá is a public-private project led by the Bank for International Settlements (BIS) and the Institute of International Finance (IIF). It tests the use of tokenized bank deposits and central bank reserves (the settlement funds that banks hold at the central bank) for cross-border payments between institutions. Because the tokens represent bank deposits and central bank reserves, they are different from stablecoins. This article sets out, in turn, what the July test confirmed and the scope that the BIS report says has not yet been tested.

Atomic settlement confirmed with real money

The BIS published its prototype report on May 27, 2026. In the accompanying press release, the BIS stated that atomic settlement, which completes a multi-step cross-border transaction on an "all-or-nothing" basis, can be achieved safely across currencies and jurisdictions. Since every linked step is either fully processed or not processed at all, a transaction does not stall partway through. The same press release also noted that tokenization does not change the legal nature of reserves and deposits or the nature of the liabilities they carry.

What sets the July test apart from the May prototype is that the tokens carried real monetary value. According to the description of real-value testing on the BIS project page, the tokens issued this time represented central bank reserves and commercial bank deposits. Six currencies were used (the won, the U.S. dollar, the euro, the yen, the pound and the Swiss franc), and 30 transactions were processed across 17 scenarios.

The use cases tested were single-currency and dual-currency payments between companies and between banks, payment-versus-payment (PvP), in which two currencies are exchanged simultaneously, and fund transfers within the same financial group. The time from payment initiation to settlement averaged about 80 seconds. This figure was obtained while the prototype was not integrated with existing large-value payment systems or banks' internal systems.

Two ledger layers and the role of each institution

The structure described in the report has two layers. One layer is the unifying ledger, which records tokenized deposits and is accessible to all participants. The other layer consists of jurisdictional ledgers, maintained separately for each jurisdiction, which record tokenized reserves. The BIS explains that this structure allows central banks to preserve autonomy over their own currencies and operations even within a shared platform.

In the real-value testing as well, authority over reserves remained with the central banks. Requests to issue tokenized reserves were initiated outside the platform, and redemption requests were approved manually by the central banks. In this test, the Bank of Korea issued, transferred and redeemed tokenized reserves on the payment instructions of the two banks. In the process, the Project Hangang platform operated by the Bank of Korea was manually linked with BOK-Wire+.

Information sharing was designed along the same lines. The report states that a shared ledger does not mean shared data, and that information is shared only when participants approve it, with privacy as the default. Anti-money laundering and sanctions screening are carried out separately and privately by each institution, and the platform ensures that all required checks have been completed before settlement. Participants share one ledger, while responsibility for issuing money and for screening stays with each institution.

Requirements outside the tested scope, as stated in the report

The May report stated that the prototype did not assess production-level requirements. The list covers cybersecurity, operational resilience, throughput and latency, liquidity optimization, failover and real-time monitoring. The report also states that testing was not carried out with transactions originating from live services or institutional systems. However, this statement was written before the July real-value testing.

The July test was also conducted within a framework different from ongoing operations. That framework consisted of contractual agreements among participating institutions, a step-by-step runbook and contingency procedures. SIX, a Swiss financial market infrastructure operator, ran the test operations and coordinated all transactions according to predefined intraday transaction windows. A two-step maker-checker process, manual approvals by banks and central banks, and multi-factor authentication for each user were features added for this test.

The legal analysis reached two conclusions. It found that settlement finality, under which a completed settlement is not reversed even in events such as the insolvency of a participant, can be achieved in all seven participating jurisdictions. However, it left the definition of technical, operational and contractual requirements suited to each jurisdiction's legal system as further work. It stated that questions such as the allocation of liability between participants and service providers, and the treatment of settlement finality and system rules in each jurisdiction, were outside the scope of the project. It also noted that differences in legal effect across jurisdictions could create operational risk when rules are enforced.

Wider participation in September and the next tests

When the May report was released, the BIS announced that the Bank of Canada had joined, and the project page now lists eight participating central banks. On September 29, the IIF announced seven new participants: Barclays, Bank of Montreal (BMO), Canadian Imperial Bank of Commerce (CIBC), Rabobank, Royal Bank of Canada (RBC), Scotiabank and Wells Fargo. On the same day, BMO, CIBC and Scotiabank each issued press releases stating that they were joining building on the results of the July real-value testing.

In its July 30 announcement, the Bank of Korea said it plans to conduct further real-value testing with transaction types and scenarios not covered this time, while the BIS stated only that testing will continue. On September 29, the IIF said that participating institutions are preparing further tests. The timing and scale of the further tests have not yet been disclosed. In the same July 30 announcement, the Bank of Korea said it would review, from various angles, linkage and interoperability between the Project Hangang platform and the Agorá platform. The example it gave was using the Project Hangang platform as Korea's jurisdictional ledger. This is still at the review stage; no decision or timeline has been set.

Three open questions for the rulebook

In the report, the BIS identified items to be analyzed for inclusion in a future rulebook (the operating rules that participants would follow in common). Among them, three questions that Korean banks and institutions will need to answer together are outlined below.

First, eligibility and the connection to the Korean ledger. The first item the BIS listed for analysis in the rulebook is participation eligibility and onboarding requirements, which include definitions of eligible institutions that differ from country to country. Since the use of the Project Hangang platform as Korea's jurisdictional ledger is under review, the issue is which Korean institutions would enter the jurisdictional ledger and the unifying ledger, and under what requirements.

Second, the point at which settlement is final. The BIS found that settlement finality can be achieved. However, which point on the unifying ledger is treated as the moment of finality, and how insolvency protection applies if a participant fails, remain tasks for the rulebook. The key question is how to align the point of finality set by Korean law with the rules of the unifying ledger.

Third, errors, cancellations and the allocation of liability. Moving to ongoing operations would require the rulebook to include procedures for handling exceptions such as errors and cancellations, standards for liability and exemptions from liability, and the allocation of liability between participants and the operator. Who sets these rules, and through what procedure, remains an open question.

The July test showed that cross-border payments can be processed with tokens carrying real monetary value, and the next task is a set of rules that participants will follow in common.